Dither is a tiny amount of noise added when you reduce bit depth, such as exporting a 24-bit mix as 16-bit. It hides the harsh distortion that rounding creates on quiet sounds. Apply it once, as the very last step, whenever audio is reduced to a lower fixed-point depth. Triangular dither is the safe default.
You finished the mix, you hit export, and your DAW asks about bit depth and dither. Check the selected export settings before bouncing. Leaving them alone is usually fine, until a quiet fade-out sounds gritty or you dither the same file twice.
This guide covers what dither does, when you need it, and which type to pick. It sits next to the full step-by-step mastering guide, and it picks up where sample rate and bit depth leaves off. There is a decision table below, plus a quick test so you can hear the difference yourself. By the end you will know exactly which export settings to choose and which to leave alone.
What dither actually does
Every digital recording stores sound as numbers, and bit depth sets how finely those numbers can describe the signal. When you drop from 24-bit to 16-bit, the extra detail has to go somewhere. Reducing bit depth maps each sample to one of the fewer values available, through rounding or truncation. This introduces quantization error.
Rounding sounds harmless, but it is not. On loud material the error is buried. On quiet material, like reverb tails, fade-outs and soft intros, the rounding error follows the music. Your ear hears it as grain, a buzzy crunch, or a tail that seems to switch off in steps.
Dither fixes this by adding a very low level of random noise before the rounding happens. The noise breaks the pattern, so the error stops tracking the music and becomes a steady, quiet hiss. A smooth hiss is much easier on the ear than distortion that comes and goes with the song.
You trade a little noise for a cleaner decay. At 16-bit that noise sits far below the music, so most listeners never notice it.
When to dither: a decision table
Compare the export depth with your DAW's processing precision. Rendering higher-precision audio to 24-bit fixed-point can also involve quantization, so even a 24-bit export may call for dither. Use triangular dither for that reduction; export 32-bit float without dither when the recipient accepts it.
Export scenario | Bit depth out | Dither? | Type |
|---|---|---|---|
Mixing in your DAW | 32-bit float internal | No | None |
Mix sent to a mastering engineer | 32-bit float | No | None |
Mix sent to a mastering engineer | 24-bit | Yes, if the DAW processes at higher precision | Triangular |
Final master for CD or a 16-bit upload | 16-bit | Yes | Triangular, or noise-shaped for quiet music |
Final master for a 24-bit upload | 24-bit | Yes, if rendered from higher-precision audio | Triangular |
Master exported to MP3 or AAC | Lossy encoder | Usually no | Dither to 16-bit first only if the encoder needs 16-bit input |
Stems for a remix | 32-bit float | No | None |
Stems for a remix | 24-bit | Yes, if rendered from higher-precision audio | Triangular |
Not sure which of these files your release needs, or when to make them? Work it through with Apollo, Orphiq's AI music strategist: ask and it builds a release timeline back from your target date.
Check your DAW's export settings and any dither-enabled plugins. Follow its documentation to avoid applying dither twice.
Which type of dither to pick
Plugins list a confusing pile of names. They come down to three ideas.
Plain random (rectangular) dither is the simplest. It can leave noise that changes with the signal. Triangular is the safer default.
Triangular dither (TPDF) is the standard choice. Its noise stays steady no matter what the music does, which makes it neutral and predictable. If you are unsure, pick this.
Noise-shaped dither moves the added noise toward frequencies where human hearing is less sensitive, usually the very top of the spectrum. The result sounds quieter and more detailed in soft passages. The total noise is higher; it just sits where you hear it less.
Type | Best for | Watch out for |
|---|---|---|
Rectangular | Rarely the right choice | Noise that shifts with the signal |
Triangular | Almost everything, loud or quiet | Slightly more audible hiss than shaped types on very quiet music |
Noise-shaped | Quiet, dynamic, acoustic or ambient music | Extra high-frequency noise if the file is processed again |
Loud, dense genres hide the difference. Quiet music shows it. A solo piano record benefits from noise shaping far more than a loud club track does.
Why dithering twice hurts
Dither is noise. Each time you add it, the noise floor rises a little. Dither a file once and the result is clean. Dither it, process it, then dither again, and you now have two layers of noise stacked on the music.
Noise shaping makes the problem worse. Further processing can make shaped noise more audible. Reserve noise-shaped dither for the final master.
Here is how it usually happens. You export a dithered 16-bit file, drop it into another session for a small fix, then export again with dither still switched on. The fix is a habit: keep intermediate files at 32-bit float when possible. For a required reduction to 24-bit fixed-point, use triangular dither. Apply dither again if the final export requires another reduction, with one dither stage per export.
One more trap: a limiter or mastering plugin with dither built in, followed by your DAW's own dither on export. Check both places and turn one off.
Hear it yourself in two minutes
You do not need to trust the theory. Test it on your own track.
Pick the quietest part of the song, such as a fade-out or a soft reverb tail.
Export that section at 16-bit with dither off.
Export the same section at 16-bit with triangular dither on.
Listen on headphones with the volume raised a little, and focus on the last second of the decay.
The undithered file tends to sound grainy as the tail dies, sometimes with a faint crackle or a stepped cutoff. The dithered file fades into a soft, even hiss. If you cannot hear a difference, it may be below the audible level in your listening setup. That does not establish whether quantization occurred.
Dither and streaming uploads
Streaming platforms do not take your file as it stands. Your distributor delivers it, and the service then encodes it for listeners. What matters on your side is what the distributor accepts and where your last bit depth drop happens.
If your distributor takes 24-bit WAV files, upload a 24-bit master, and dither once if creating it reduced your DAW's processing precision. Accepting 24-bit does not tell you whether quantization happened when the master was made. If it requires 16-bit, dither once on the way to 16-bit and send that file.
Do not dither a file because it is "going to streaming." Dither because the bit depth is dropping. Keep loudness and true peak headroom in mind too, which is covered in the mastering for streaming guide, and see the mastering guide for where dither fits in the chain. More guides like this one for artists sit in the same library.
Frequently Asked Questions
Do I need to dither if I export at 24-bit?
Possibly. Dither matters whenever quantization happens, for example from 24-bit to 16-bit, or from your DAW's higher-precision audio to 24-bit fixed-point. Exporting 32-bit float needs none.
Can I hear dither?
Rarely. It sounds like a faint, steady hiss far below the music. Its job is to remove the grainy distortion on quiet sounds, so a good result is one you stop noticing.
Is noise-shaped dither better than triangular?
It depends on the music. Noise shaping helps quiet, dynamic tracks by moving noise to less audible frequencies. Triangular is the safer default, especially if a file may be processed again.
Should I dither stems or mixes sent to mastering?
Send 32-bit float with no dither if the recipient accepts it. For 24-bit files, use triangular dither once. The mastering engineer will dither once at the final export.
Read Next
How to Plan a Music Release: Step-by-Step Checklist is the next step when you're ready to put your finished master to work.